[Tex/LaTex] Using package minted with non-english characters

Question

I would like to use LaTeX package minted to colorize my c and Cuda code. I need to write comments in the code in czech. I use package inputenc with utf-8 and czech characters (ěščřžýáíé) work good in whole paper but minted block:

LaTeX source:

    \begin{minted}{c}
        double m[4];    // staticky alokované pole
    \end{minted}

Output pdf:

    double m[4];    // staticky alokovan\unhbox \voidb@x \bgroup \let \unhbox \voidb@x \

EDIT:
I have tried XeTeX. minted works fine with XeTeX and czech characters. But assymptote doesn't work for me with XeTeX. asy command used on .asy file from XeTeX utilize 100% of my cpu but not generate any output in casual time.

EDIT:
Example for pdflatex:

   \documentclass[a4paper, pt11]{article}
   \usepackage[utf8]{inputenc}
   \usepackage{fullpage}
   \usepackage[english, czech]{babel}

   % More symbols
   \usepackage{amsmath}
   \usepackage{amssymb}
   \usepackage{latexsym}


   \usepackage{graphicx}
   \usepackage{asymptote}
   \usepackage{minted}

   \begin{document}
   \begin{figure}[h]
    \begin{asy}
        import unicode;
        size(15cm);
        fill((0, 0)--(0.5, 0)--(0.5, 1)--(0, 1)--cycle, red);
        fill((0.5, 0)--(6, 0)--(6, 1)--(0.5, 1)--cycle, blue);
        fill((6, 0)--(32, 0)--(32, 1)--(6, 1)--cycle, green);
        draw((0.25, 1.1)--(0.25, 1.9));
        label("znaménko", (0.25,2), N);
        label("exponent", (3.25,1), N);
        label("mantisa", (18,1), N);
        for(int i=0; i < 64; ++i) {
            draw((0.5i, 0)--(0.5i + 0.5, 0)--(0.5i + 0.5, 1)--(0.5i, 1)--cycle);
        }
        label("$63$",(0.25,0),S);
        label("$52$",(5.75,0),S);
        label("$0$",(31.75,0),S);
    \end{asy}
    \caption{Uspořádání \emph{double} v paměti}
    \label{doublemem}
   \end{figure}

    \begin{minted}{c}
    double m[4];    // staticky alokované pole
    m[0] = 1;       // přímé nastavení první složky
    double *n = (double *) malloc(pocet_cisel * 4 * sizeof (double));   // dynamická alokace
    n[4 + 0] = 1;   // přímý přístup do první složky druhého čísla
    free(n);        // vyčištění paměti
   \end{minted}
   \end{document}

Answer 1

Here's a patch for fancyvrb and minted. This should let minted handle whatever Unicode pdfTeX can handle. @KonradRudolph All of this comes down to making sure everything is processed by \detokenize before it is saved to file.

\documentclass[11pt]{article}
\usepackage[utf8]{inputenc}
\usepackage[english, czech]{babel}

\usepackage{fullpage}
\usepackage{minted}

\makeatletter
\newcommand{\minted@write@detok}[1]{%
  \immediate\write\FV@OutFile{\detokenize{#1}}}%

\newcommand{\minted@FVB@VerbatimOut}[1]{%
  \@bsphack
  \begingroup
    \FV@UseKeyValues
    \FV@DefineWhiteSpace
    \def\FV@Space{\space}%
    \FV@DefineTabOut
    %\def\FV@ProcessLine{\immediate\write\FV@OutFile}% %Old, non-Unicode version
    \let\FV@ProcessLine\minted@write@detok %Patch for Unicode
    \immediate\openout\FV@OutFile #1\relax
    \let\FV@FontScanPrep\relax
%% DG/SR modification begin - May. 18, 1998 (to avoid problems with ligatures)
    \let\@noligs\relax
%% DG/SR modification end
    \FV@Scan}
    \let\FVB@VerbatimOut\minted@FVB@VerbatimOut

\renewcommand\minted@savecode[1]{
  \immediate\openout\minted@code\jobname.pyg
  \immediate\write\minted@code{\expandafter\detokenize\expandafter{#1}}%
  \immediate\closeout\minted@code}
\makeatother

\begin{document}

\begin{minted}{c}
double m[4];    // staticky alokované pole
m[0] = 1;       // prímé nastavení první složky
double *n = (double *) malloc(pocet_cisel * 4 * sizeof (double));   // dynamická alokace
n[4 + 0] = 1;   // prímý prístup do první složky druhého císla
free(n);        // vycištení pameti
\end{minted}

\mint{c}|double m[4];    // staticky alokované pole|

\end{document}

Edit 2013/07/30

Equivalent code is now incorporated into the minted development version on GitHub.